Inauthentic social media activity sought to sow distrust in midterm elections, report finds

In the final weeks before the midterm elections, there was a proliferation of inauthentic activity on social media that sought to sow distrust in the election results, according to a new report.An AI-powered software company called Cyabra maintains that the inauthentic activity did not change the outcome of any of the five elections listed in its report: the gubernatorial races in Ohio and Georgia and the Senate races in Michigan, Maine, and Kansas. Rather, the company’s concern rested primarily in how narrative attacks shaped public perception regarding these races, which, in turn, affect how journalists or pollsters may evaluate voter sentiment.“The target was trust in the result, not any single race,” the final report states. “Both parties were hit, the activity on X peaked on Election Day, and the most-viewed fake content was fabricated results posted before polls closed.” Cyabra estimates that 18.7% of X posts about the five races on Election Day originated from inauthentic accounts, most of which were “sleeper” accounts. In short, these accounts woke up in the final stretch of the midterm elections. Cyabra says 84% of the 8,344 inauthentic accounts that were active between Oct. 20 and Nov. 3 were sleepers, meaning they had not appeared in its analysis between Aug. 31 and Oct. 19.Most of the sleeper accounts with a known creation date had already existed before Oct. 1. Only 76 sleeper accounts were created after that date.Regarding who was behind these inauthentic accounts, Cyabra concludes that nearly 85% of these accounts with known locations were based in the United States. The company says its analysis does not rule out foreign involvement, but it does not attribute the nefarious activity to any foreign country or actor either.Much of this activity was concentrated on X, far more than Facebook and Instagram. TikTok was not included in the report.Another concerning trend discovered by Cyabra was that fabricated results in Michigan and Ohio spread hours before polling sites closed in those states.An inauthentic account posted fake results in the Michigan Senate race at 1:35 p.m. last Tuesday and asked Grok to “confirm this.” Grok’s response was not featured in the report, as Cyabra was more concerned with the “perceived neutrality of AI” lending credibility to a fake claim, regardless of the chatbot’s answer.Notably, the same bogus results were shared by an authentic account at 6:25 p.m. before votes were officially counted. A similar trend was seen in the Ohio gubernatorial race.With less than two years before the next presidential election, Cyabra anticipates the swarm of inauthentic accounts to become more sophisticated.“In 2028, Cyabra expects the same playbook, faster and closer to the count: fresh sleeper accounts each cycle, AI chatbots invoked as a credibility stamp, and the same accounts working many races at once,” the report reads.FRANCE BLAMES IRAN FOR SUPERMARKET REFRIGERATOR CYBERATTACKCyabra is raising the alarm now so that election officials are aware of the issue and can start working to mitigate it. The company recommends that governments at the federal, state, and local levels each address this inauthentic activity on social media in their own way while urging voters to remember that election results do not come from posts or chatbots.Cyabra, founded in 2017 by veterans of Israeli military intelligence units, uses artificial intelligence to analyze data that helps global enterprises, media organizations, and government agencies detect disinformation, fake profiles, bots, deepfakes, and other coordinated influence operations.